Github user jose-torres commented on the issue:
https://github.com/apache/spark/pull/20726
I think both failures don't reflect a deeper problem:
* The first was because columnar batch scans silently shortcutted around
the readerFactories lazy val, and would error out if it were called. I fixed
this, but the resulting code is kinda messy, so let me know if you see a better
way to do it.
* The second was a bit of fragile logic, basically measuring which
streaming metric interval every data source method gets called in. The change
does indeed move reader factory creation from addBatch to queryPlanning; this
is expected and non-problematic.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
